As our first heatsink review roundup was being prepared, it became evident that the explanation of the testing methods was much too extensive to remain in the review; thus, this article was born. Our method is unique in that it ensures the heatsink is the only variable; all other elements, including the fan, are kept consistent and controlled in all the heatsink tests. We think you'll appreciate the strengths of this approach. Here is the lowdown on our Unique Heatsink Testing Methodology.

Just what the doctor ordered, excellent

This sounds like a pretty good measure of heatsink quality. It's certainly excellent to have the heatsink reviewed with different voltage Panaflos rather than Delta screamers. I have just a few questions:

1) Why not use a CPU simulating heatblock like, for example, Tom's Hardware Guide? By which I'm also asking, what do you think of this approach as compared to "real-world" tests?

2) How did you decide on placing the mic so close to the fan? I suppose as long as you're consistent it doesn't matter, but I'm wondering if maybe a larger distance might help reduce measurement error.

3) I suppose you'll just be ignoring 60mm heatsinks, since pretty much every top-flight heatsink now accepts 80mm fans. Yes/No?

1) $$ & time & usefulness. First two are simple. The third is also simple: when it's no longer used what function will it have? I can recycle/reuse a mobo+cpu anytime -- even in between HS testing -- but a heatblock? it's garbage that can't even be recycled. Also it's no small challenge to duplicate the mounting system so it's just like a mobo -- and the clip/mounting system, as mentioned in the article, IS critical, a key part of the whole design. Sorta like tires/wheels on cars.

2) It's called "nearfield" sound measurement -- it's what Storage Review does. Used when you don't have a highly sensitive meter that lets you measure from farther away, or when ambient noise is too high. No way most SLMs will register ANYTHING with the Panaflo at 5V from 1 meter away.

3) The whole point is to use only one fan, as much as possible. Then you know w/little doubt that HS A is better than HS B; we're not comparing a lousy HS + KingKong fan vs. a Godzilla HS + 5V Panaflo. You can then choose the HS on the basis of its intrinsic cooling power with low airflow, quiet fans. 60s are too loud anyway.

Look for the first roundup in the next day or two.

excellent! Finally the myths about low airflow performance will be dispelled, or at least they won't be myths anymore.

One other thing to try would be to run each HS passive with a standard undervolt/underclock. That might be of use to absolute silence seekers though I admit it would be harder to test due to the risks to equipment.

Thank you Mike for what will likely be the first true comparison of heatsinks on the WWW. So far, my many many hours of HSF review surfing have been mostly an exersize in frustration, reading between the lines, and guessing. Trying to find comparisons where the stock fan comes off and a common reference performance fan is used. Some have come close, but none have really answered the real questions. So... Thank you Thank you Thank you.

I would suggest putting this point first:

Is it considered so good that we just have to see how it does with a quiet fan?

This point alone will appeal to anyone with pride in their product. I also suspect that it is the very best coolers (all Cu, big, well made, etc.) that will do best when run quiet. It seems these are seldom thought of for quiet applications except by the likes of us. Like the Thermalright SLK 800, which fits your other criteria, but the 'so good' factor is the most appealing appeal.

Also, I have often wondered about some HS designs that might require ducting. That is, some heatsinks might need a 80mm -> 60mm converter. Should an open or closed sides type be used? Some HS's could impede airflow and benefiet from the open sides duct, others might better use the increased focus of a closed duct. I run a cheap Arkua 6129 (no Cu core ), tossed the 60mm loud fan, made a closed 80>60 duct with some heavy synthetic felt and hotmelt glue, then glued the duct to the plastic 60mm fan mount that is screwed to the extrusion. I am waiting for my Panaflos, but my 80mm Thermaltake thermistor fan runs real slow and cools much better/quieter than the original 60mm. The Thermaltake doesn't mind the constriction. My rig works better than it should given everything against it, but a top end Arkua or Thermal Integration might do really quite well under a Panaflo (and cheap ). Still the question remains, with a Panaflo, especially undervolted, how would the restricted airflow interplay with performance (stall effects etc.). Empirical testing will tell all, but some very intersting suprises might be lurking in the results, especially at 5v. For unducted HS's, I even wonder if standoffs (let the air leak and move) might help at 5v. Sorry for all the questions.

Im glad to see a scientific approach to heatsink testing in a quiet enviroment. The resulting socket-a HS article was great--thanks. I am curious about one thing: How does ambient temperature affect the dissapation characteristics of the HS's? Do some perform better in warmer enviroments than others? I imagine that there are some differences in how the heatsinks perform in varying ambient temperatures, so I wonder--would it be better to to test the heatsinks in an environment more like the inside of a computer case, where the temps would be much warmer than room temps?

AFAIK, temp is additive. If the ambient is 30C, naturally, the CPU & HS are bound to start at this temp. I can't imagine how or why any HS would work better in a warmer environment; if anyone know different, please post here.

The question of whether the testing should be done in a case:

I don't think it really matters if you consider my first point above. I make the point in the review that the single most important piece of data about any of the HS is "Temp Rise" -- or temp rise above ambient. That takes away the issue of the ambient temp. You should be able to predict (more or less) the temp of any of the tested HS with a similar heat source (CPU) by adding the ~case temp around the CPU area (or fan intake) to the "Temp Rise" figure.

The C/W figure gives you a performance indicator independent of the heat source, but it assumes that the 65W figure used is correct. Even though that is based on AMD specs, I have no way to verify this on my particular XP1600+ sample. Also I don't KNOW that Prime95 is pushing the CPU to 100%.

BTW -- re a comment someone mentioned else about BURNK7 (CPUBURN)causing higher CPU temps than Prime95. After seeing that post I downloaded the friendly GUI windows version and ran it at high priority & compared it to Prime95. I did not find any difference in the max diode temps. The only significant difference seemed to be that Prime95 causes the temp to waver slightly from the max down to maybe -2C below, while once the max temp is reached with CPUBURN, it stays there rock solid. So I will adopt CPUBURN for the next set of HS tests, as the temp data is the same; the lack of waivering makes it easier to determine when max temp is reached.

Multiple testing of HS inside a case is an incredible pain; I tried it a few times & gave up because of the physical hassles. Bigger chances of mucking things up. Also, every case different anyway, so that doesn't always help.

Great work on the heatsink testing. I am finally seeing someone test the heatsinks without the m/b sensor.

You also point out a VERY IMPORTANT point, temp rise. I wondered around many messageboards, seems like many users only concern the cpu temp and failed to realize that chassis ambient would affect the cpu temp as well.

Just a quick question, where do you place the ambient sensor?
I am wondering if there could be a slight temp difference between the temp at the sensor and temp right above the cpu intake fan.

As far as driving cpu to 100% power, you can try to get a software called "maxthk7f.exe" from AMD. This software will allow you to drive the AMD processor to 100% power.

Again, great work.

Intel has a software called "P4maxpower6.exe". This software will allow you to set the throttle on the p4 chip. You can basically run the chip at 100%, 90%, 75%....or whatever you feel like. Again, you will have to get it from Intel.

Yes there is, but this is tricky.

The temp directly above the fan intake at the beginning of a 20~30 min burn is about the same as 6" above, which is where I measure the ambient. But by the end, it is higher by a few degrees. Does this mean my ambient is too low? Maybe.

The thing is that the air immediately around the CPU/HS is getting warmed by the CPU as well as other components on the board. You COULD say this is all part and parcel of the ambient, but as it is a direct result of the CPU being pushed, I think otherwise. Also, this means the ambient is actually changing in the course of the test, which really complicates things. And the tests are complicated enough!

For simplicity and repeatability, I chose the 6" distance. It is little affected by the warming of the air immediately around the HSF and stays stable through the course of the test: ~1 degree variance. If I was running the test inside a case, I might take an average of several temp readings before each test: at front low case intake vent, near CPU/HSF, and one around the PCI slots -- then average them.

Thanks for the links to the other burn programs. Will check them out when I have time.